Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754455AbcDJUHt (ORCPT ); Sun, 10 Apr 2016 16:07:49 -0400 Received: from mail.linuxfoundation.org ([140.211.169.12]:53187 "EHLO mail.linuxfoundation.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1759597AbcDJS6r (ORCPT ); Sun, 10 Apr 2016 14:58:47 -0400 From: Greg Kroah-Hartman To: linux-kernel@vger.kernel.org Cc: Greg Kroah-Hartman , stable@vger.kernel.org, Raghava Aditya Renukunta , Johannes Thumshirn , Tomas Henzl , "Martin K. Petersen" Subject: [PATCH 4.4 039/210] aacraid: Fix RRQ overload Date: Sun, 10 Apr 2016 11:34:20 -0700 Message-Id: <20160410183528.053514618@linuxfoundation.org> X-Mailer: git-send-email 2.8.0 In-Reply-To: <20160410183526.651820045@linuxfoundation.org> References: <20160410183526.651820045@linuxfoundation.org> User-Agent: quilt/0.64 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 4920 Lines: 152 4.4-stable review patch. If anyone has any objections, please let me know. ------------------ From: Raghava Aditya Renukunta commit 3f4ce057d51a9c0ed9b01ba693df685d230ffcae upstream. The driver utilizes an array of atomic variables to keep track of IO submissions to each vector. To submit an IO multiple threads iterate through the array to find a vector which has empty slots to send an IO. The reading and updating of the variable is not atomic, causing race conditions when a thread uses a full vector to submit an IO. Fixed by mapping each FIB to a vector, the submission path then uses said vector to submit IO thereby removing the possibly of a race condition.The vector assignment is started from 1 since vector 0 is reserved for the use of AIF management FIBS.If the number of MSIx vectors is 1 (MSI or INTx mode) then all the fibs are allocated to vector 0. Fixes: 495c0217 "aacraid: MSI-x support" Signed-off-by: Raghava Aditya Renukunta Reviewed-by: Johannes Thumshirn Reviewed-by: Tomas Henzl Signed-off-by: Martin K. Petersen Signed-off-by: Greg Kroah-Hartman --- drivers/scsi/aacraid/aacraid.h | 2 ++ drivers/scsi/aacraid/commsup.c | 28 ++++++++++++++++++++++++++++ drivers/scsi/aacraid/src.c | 30 +++++++----------------------- 3 files changed, 37 insertions(+), 23 deletions(-) --- a/drivers/scsi/aacraid/aacraid.h +++ b/drivers/scsi/aacraid/aacraid.h @@ -944,6 +944,7 @@ struct fib { */ struct list_head fiblink; void *data; + u32 vector_no; struct hw_fib *hw_fib_va; /* Actual shared object */ dma_addr_t hw_fib_pa; /* physical address of hw_fib*/ }; @@ -2113,6 +2114,7 @@ static inline unsigned int cap_to_cyls(s int aac_acquire_irq(struct aac_dev *dev); void aac_free_irq(struct aac_dev *dev); const char *aac_driverinfo(struct Scsi_Host *); +void aac_fib_vector_assign(struct aac_dev *dev); struct fib *aac_fib_alloc(struct aac_dev *dev); int aac_fib_setup(struct aac_dev *dev); void aac_fib_map_free(struct aac_dev *dev); --- a/drivers/scsi/aacraid/commsup.c +++ b/drivers/scsi/aacraid/commsup.c @@ -90,6 +90,28 @@ void aac_fib_map_free(struct aac_dev *de dev->hw_fib_pa = 0; } +void aac_fib_vector_assign(struct aac_dev *dev) +{ + u32 i = 0; + u32 vector = 1; + struct fib *fibptr = NULL; + + for (i = 0, fibptr = &dev->fibs[i]; + i < (dev->scsi_host_ptr->can_queue + AAC_NUM_MGT_FIB); + i++, fibptr++) { + if ((dev->max_msix == 1) || + (i > ((dev->scsi_host_ptr->can_queue + AAC_NUM_MGT_FIB - 1) + - dev->vector_cap))) { + fibptr->vector_no = 0; + } else { + fibptr->vector_no = vector; + vector++; + if (vector == dev->max_msix) + vector = 1; + } + } +} + /** * aac_fib_setup - setup the fibs * @dev: Adapter to set up @@ -151,6 +173,12 @@ int aac_fib_setup(struct aac_dev * dev) hw_fib_pa = hw_fib_pa + dev->max_fib_size + sizeof(struct aac_fib_xporthdr); } + + /* + *Assign vector numbers to fibs + */ + aac_fib_vector_assign(dev); + /* * Add the fib chain to the free list */ --- a/drivers/scsi/aacraid/src.c +++ b/drivers/scsi/aacraid/src.c @@ -156,8 +156,8 @@ static irqreturn_t aac_src_intr_message( break; if (dev->msi_enabled && dev->max_msix > 1) atomic_dec(&dev->rrq_outstanding[vector_no]); - aac_intr_normal(dev, handle-1, 0, isFastResponse, NULL); dev->host_rrq[index++] = 0; + aac_intr_normal(dev, handle-1, 0, isFastResponse, NULL); if (index == (vector_no + 1) * dev->vector_cap) index = vector_no * dev->vector_cap; dev->host_rrq_idx[vector_no] = index; @@ -452,36 +452,20 @@ static int aac_src_deliver_message(struc #endif u16 hdr_size = le16_to_cpu(fib->hw_fib_va->header.Size); + u16 vector_no; atomic_inc(&q->numpending); if (dev->msi_enabled && fib->hw_fib_va->header.Command != AifRequest && dev->max_msix > 1) { - u_int16_t vector_no, first_choice = 0xffff; - - vector_no = dev->fibs_pushed_no % dev->max_msix; - do { - vector_no += 1; - if (vector_no == dev->max_msix) - vector_no = 1; - if (atomic_read(&dev->rrq_outstanding[vector_no]) < - dev->vector_cap) - break; - if (0xffff == first_choice) - first_choice = vector_no; - else if (vector_no == first_choice) - break; - } while (1); - if (vector_no == first_choice) - vector_no = 0; - atomic_inc(&dev->rrq_outstanding[vector_no]); - if (dev->fibs_pushed_no == 0xffffffff) - dev->fibs_pushed_no = 0; - else - dev->fibs_pushed_no++; + vector_no = fib->vector_no; fib->hw_fib_va->header.Handle += (vector_no << 16); + } else { + vector_no = 0; } + atomic_inc(&dev->rrq_outstanding[vector_no]); + if (dev->comm_interface == AAC_COMM_MESSAGE_TYPE2) { /* Calculate the amount to the fibsize bits */ fibsize = (hdr_size + 127) / 128 - 1;